Abstract: This paper puts forward to the concept of similarity between member and member set，proves and analyses some useful propositions and corollaries，gives an if and only if condition which can hold the coherence of cluster set when a member joined to the cluster set，and describes a heuristic iterative algorithm for group clustering.The result shows that the algorithm has better clustering performance and lower coherent deviation index.

Key words: group clustering, similarity, coherence, deviation

摘要： 提出了成员与成员集的相似度概念，证明与分析了一些有用的命题和推论，给出了一个保持聚集一致性的成员与聚集相聚的充分必要条件，描述了一个群体聚类的启发式迭代算法。实例测试表明，提出的算法有较好的聚类性能和较低的一致性偏差指标。

关键词: 群体聚类, 相似度, 一致性, 偏差